Resource Use by Hospital Type for Frequent Inpatient Pediatric Conditions

This was a cross-sectional, observational analysis of pediatric hospitalizations for frequent inpatient conditions in the United States using the 2016 Healthcare Cost and Utilization Project (HCUP) Kids’ Inpatient Database (KID), which is the most current available.¹¹  We included all discharges for patients <18 years of age with a primary diagnosis of pneumonia, asthma, acute bronchiolitis, asthma, mood disorders, appendicitis, epilepsy, skin and soft tissue infections (SSTI), or fluid and electrolyte disorders, identified by HCUP Clinical Classification Software Refined, and we excluded hospitalizations for neonates identified by All Patient Refined Diagnostic Related Groups.⁴  Clinical Classifications Software Refined is based on the International Classification of Diseases, 10th Revision, Clinical Modification and collapses related diagnosis codes into the principal category. The included frequent inpatient conditions were chosen on the basis of the acute care conditions most frequently accounting for pediatric hospitalizations within KID.¹² 

KID is maintained by the Agency for Healthcare Research and Quality.¹¹  It includes data from CHs and non-CHs, patient demographics, hospital characteristics, diagnoses, lengths of stay (LOSs), and charges.¹¹  The institutional review board at Baylor College of Medicine determined this deidentified data did not constitute human subjects research.

Primary outcomes were LOS and total cost of hospitalization. LOS was determined by subtracting the admission from the discharge date and was measured in days, and the minimum LOS was 1 day, even if admission and discharge occurred on the same calendar day. Hospital charge was the total amount charged by the hospital. The HCUP cost-to-charge ratio file provided hospital-specific ratios, allowing for the conversion of charges to estimated costs.

The primary independent variable was hospital type, categorized as nonchildren’s, nonteaching (NCNT); nonchildren’s, teaching (NCT); and freestanding CHs, as classified by the American Hospital Association.¹¹  Patient and hospital characteristics consisted of age, race or ethnicity, sex, insurance type, Hospitalization Resource Intensity Scores for Kids (H-RISK)¹³  severity of illness, median zip code–associated income, hospital bed size, and number of complex chronic conditions (CCCs).¹⁴ 

Analyses were performed by using SAS 9.4 (SAS Institute, Inc, Cary, NC). Because of skewed distribution of data toward higher LOS and cost before and after log transformation, median values were examined in unadjusted analysis. A Rao-Scott χ² test was used to compare categorical variables, and a Kruskal-Wallis test was used to compare medians of continuous variables across hospital types. Survey-weighted multivariable linear regression with log transformation was used to assess differences in mean LOS and costs, controlling for patient and hospital characteristics. LOS was controlled for in-cost analysis. We considered each hospital as a cluster for the purpose of calculating SEs. Results were reported as estimates with P values (P < .05 was considered significant).

In 2016, there were 354 456 pediatric discharges for frequent inpatient conditions (Table 1). NCT hospitals were the most frequent discharging hospital type (56%). Non-CHs combined (NCNT and NCT) cared for >70% of frequent inpatient condition discharges. Payer-mix distribution was similar, with nearly 60% of discharges attributable to public payers within each hospital type. NCT hospitals and CHs cared for a higher H-RISK severity and a greater proportion of children with CCCs. Approximately one-third of discharges for frequent inpatient conditions in CHs had ≥2 CCCs. Pneumonia accounted for the highest proportion of discharges at NCNT hospitals compared with acute bronchiolitis within NCT hospitals and CHs (Table 2).

In unadjusted analysis, the median LOS was 2.00 days at all hospital types with interquartile range (IQR): 1.00–4.00 at CH compared with IQR: 1.00–3.00 at NCT and NCNT hospitals. Median total costs at CHs were $6298 (IQR: $3643–$11 468) compared with $3725 (IQR: $2252–$6583) at NCT hospitals. After controlling for patient and hospital factors, discharges for frequent inpatient conditions at CHs had 0.48% longer mean LOSs and 61% greater costs compared with NCTs (P < .01). LOS and costs for frequent inpatient conditions at NCNT were not significantly different compared with NCT. Similar findings were seen for each of the individual conditions, with CHs having 28% to 108% significantly higher costs than NCT hospitals (Table 3).

In this pediatric study for frequent inpatient conditions, we found a 61% higher cost for frequent inpatient conditions in CH compared with NCT hospitals in adjusted analysis. Our findings are consistent with previous findings¹⁰  and further the understanding of resource use using costs rather than charges (because costs more closely approximate the monetary value of the care rather than what hospitals billed for services). There are multiple potential explanations for the differences in costs between CHs and NCT hospitals. First, it is possible that CHs provide higher cost, specialized care that is not commonly available at NCT hospitals. In our study, we found the highest H-RISK score and a significant proportion of discharges with ≥2 CCCs in CHs. Current literature reveals CHs continue to serve as a safety net for specialized and vulnerable populations. In recent work by Lopez et al⁴  the authors demonstrated that CHs cared for a disproportionate volume, cost, and diversity of conditions compared with general hospitals. CHs also cared for the majority of nonneonatal high-cost hospitalizations across all hospitals, including those requiring ventilator support or tracheostomy with ventilatory support for >96 hours with extensive procedure or extracorporeal membrane oxygenation.⁴  Thus, it is possible that we still lack sufficient measures to control for patient-level factors and that higher costs are warranted for the specialized care at CHs.

An alternative is that CHs are more costly in delivering standard care for frequent inpatient conditions compared with NCT hospitals. CHs may have a higher intensity of care and services, potentially related to academic mission, including more subspecialty consultations, diagnostic testing, and broad spectrum antimicrobial therapy.^15,16  In complicated pneumonia, hospital-specific differences in clinical management have been revealed to be a key factor in hospitalization cost variations.¹⁷  Without data on patient outcomes, it is difficult to interpret whether high intensity of services and higher cost of care at CH is necessary or whether it is a result of overuse (unnecessary care resulting in higher resource use).^18,19  Within pediatric hospitalizations at CH, recent data demonstrate significant variability in the use of non–evidence-based testing and treatment of frequent inpatient conditions such as asthma and bronchiolitis.²⁰  Parikh et al²⁰  established use targets across CHs to decrease the use of nonrecommended management for hospitalized children with frequent inpatient conditions. On the basis of criteria established in evidence-based guidelines, diagnostic testing and treatment rates at top-performing CHs were used to establish achievable benchmarks for their peers.²⁰  Expansion of this work to NCNT and NCT hospitals could help inform practice and establish new targets, standardize inpatient care, and support judicious resource use for frequent inpatient conditions across all hospital types.

There were methodologic limitations to our study. The data within KID do not include observation status stays, and we do not know how this may impact resource-use comparisons. It was also not possible to compare outcomes such as readmissions across hospital types because of limitations of KID. Additionally, cost-to-charge ratios were used, but these calculations may not fully estimate the true cost, particularly in non-CHs, where cost-shifting may occur to adult hospitalizations, and results must be interpreted with the understanding that costs accrued by the hospital may not reflect what hospitals are actually paid for services. Our study has the limitations inherent to using administrative data sources, including coding errors, missing variables, and possible misclassifications of hospital types. Additionally, because of database limitations, in our study, we specifically compare freestanding CHs with other hospital types, and nonfreestanding CHs were included in the NCT category.

CHs revealed higher estimated costs in caring for frequent inpatient conditions despite controlling for patient- and hospital-level factors. Further research is warranted to explore whether we lack sufficient measures to control for patient-level factors and whether higher costs are justified by the specialized care at CHs.

Future efforts should be made to continue to explore differences in care across all hospitals that treat children.